Clinical Perspectives on the Use of Computer Vision in Glaucoma Screening

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Glaucoma is a leading cause of irreversible blindness globally, necessitating early detection.
• Current diagnostic methods for glaucomatous optic neuropathy (GON) face challenges due to variations in optic nerve head presentation, defining diagnostic limits, and specialist subjectivity.
• Advances in retinal imaging and functional assessments exist, but objective references for GON detection are still needed.

Purpose of the Study:

• To analyze the application of deep learning (DL) methodologies in glaucoma screening.
• To evaluate DL's potential in reducing GON assessment costs and specialist workload.
• To improve the speed and consistency of glaucoma diagnosis.

Main Methods:

• Utilized deep learning (DL) methodologies for analyzing various diagnostic data, including color fundus photographs (CFPs), visual field (VF) tests, and optical coherence tomography (OCT) scans.
• Compared DL-based screening with traditional clinical assessments.
• Investigated the minimum detection limits of GON characteristics using these methodologies.

Main Results:

• Deep learning (DL) methodologies have demonstrated success in assisting the diagnosis and progression monitoring of glaucomatous optic neuropathy (GON).
• DL offers objective classification, potentially mitigating biases in expert decisions.
• Automated glaucoma screening using DL yields robust results that closely align with clinical reality.

Conclusions:

• Deep learning (DL) methodologies are effective tools for automated glaucoma screening.
• DL can enhance the objectivity, speed, and consistency of diagnosing glaucomatous optic neuropathy (GON).
• Implementing DL in glaucoma screening can lead to more reliable and cost-effective assessments.